Understanding microwave heating for materials synthesis

1. The dielectric terms

A material in an alternating electric field polarises. The response is one complex number:

ε′ — dielectric constant (real part)

Polarisation in phase with the field. Energy stored and returned each cycle. Sets surface reflection and the wavelength inside the material.

ε″ — loss factor (imaginary part)

Polarisation lagging the field by 90°. Energy dissipated as heat: P_v = ωε₀ε″E_rms². This, not tan δ, sets the heating rate for a given internal field.

tan δ — loss tangent

Energy lost per cycle relative to energy stored. A convenient dimensionless ranking, but it is a ratio: a material can have a high tan δ and still heat slowly if ε″ is small.

ε″_eff — effective loss factor

At one frequency you cannot separate conduction loss from polarisation loss, so they are lumped: ε″_eff = ε″_d + σ/(ωε₀). Tabulated "ε″" almost always means this.

Lossy insulator

An insulator has few free charges, so it carries little DC current; its bound charges polarise instead. If that polarisation cannot follow the oscillating field exactly, it lags, and part of the field energy is dissipated each cycle. A dielectric with significant ε″ is a lossy insulator. At microwave frequencies materials fall into three groups:

Where the polarisation comes from

• Electronic — electron cloud shifts. Follows the field to UV. Lossless at GHz.
• Ionic — cations and anions displace. Resonates in the far-IR. Contributes to ε′ of ceramics; its tail is the small intrinsic loss of good ceramics.
• Dipolar — permanent dipoles rotate. Relaxes at GHz in small-molecule liquids. The main loss route for polar solvents, and for polar domains in ferroelectrics.
• Interfacial (Maxwell–Wagner) — charge piles up at boundaries between regions of different conductivity: grains, pores, phases. Important in powders and compacts.

Each mechanism adds to ε′ below its characteristic frequency and drops out above it. ε″ peaks where it is lagging — around that frequency.

2. Relaxation and conduction

Dipolar loss follows the Debye relaxation, with one relaxation time τ — the time a dipole takes to reorient. Add a conduction term for mobile charge:

• Dipolar ε″ peaks at ωτ = 1, with height Δε/2. ε′ falls through the same region, from ε_s to ε_∞.
• Conduction loss scales as 1/f — it dominates at low frequency, and in anything with mobile ions or carriers.
• Why a large ε′ does not mean a large tan δ. A big ε′ means a big Δε, so a big possible ε″ — at most Δε/2. Whether you get it depends on τ. Water has the larger ε′ but relaxes at 19 GHz, far above 2.45 GHz, so it sits low on the rising flank. Ethanol has a third of the ε′ but relaxes at ~1 GHz, right next door, so it couples better.

Drag τ and watch the peak sweep past 2.45 GHz — that is the whole story of why materials couple or do not. Drag σ up and the 1/f conduction tail takes over from the left.

3. Why the properties change with temperature

Three quantities in those equations are temperature dependent:

• τ falls as T rises. Lower viscosity and weaker hydrogen bonding let dipoles turn faster, roughly Arrhenius. The loss peak moves to higher frequency.
• Δε falls as T rises. Thermal agitation fights alignment, so ε_s drops, roughly as 1/T. Ferroelectrics are the exception — ε_s spikes at the Curie point.
• σ rises as T rises, exponentially: σ = σ₀ exp(−E_a/kT). Ionic mobility and carrier count both increase.

At a fixed 2.45 GHz, which way ε″ moves depends on where the material sits relative to its own loss peak. There are three characteristic behaviours, and every material is one of them:

• ωτ < 1 → self-limiting. Water's peak is at 19 GHz, well above 2.45 GHz. Heating pushes it further away, so ε″ drops — by a factor of 11 from 0 to 95 °C. Hot water absorbs far less than cold water, which is why a microwave heats a mug unevenly and why runaway is not a problem in aqueous work.
• ωτ > 1 → rises, then falls. Glycerol starts viscous with its peak below 2.45 GHz. Heating sweeps the peak up through the operating frequency, so ε″ rises to a maximum and then falls away. Ice behaves the same way on melting, which is why frozen food heats erratically.
• Conduction-dominated → runaway. Fe₂O₃, and oxide ceramics generally, gain conductivity exponentially with temperature. ε″ climbs by three orders of magnitude between 25 and 1000 °C. Hotter means more absorbing means hotter: local hot spots, cracking, and melting at the centre of a nominally uniform pellet.

In most solids ε′ creeps up slowly with T through lattice expansion. The dramatic changes are all in ε″.

4. Penetration depth

A wave entering a lossy material decays as it travels. Field amplitude falls as e^−αz, power as e^−2αz:

• D_p ∝ 1/f. Higher frequency heats a thinner surface layer. At 24 GHz water's D_p is about 1 mm rather than 2 cm.
• d ≫ D_p — heat lands near the front face and the interior waits on conduction. d ≪ D_p — heating is nearly uniform, but most of the power goes straight through (in a closed cavity it comes back round for another pass).
• It collapses on heating. Alumina's D_p is 61 m at 25 °C and a few centimetres by 1200 °C. This is why a cold ceramic will not couple at all and then, once a susceptor has got it hot, suddenly absorbs strongly. The switch can be abrupt enough to crack the piece.
• High ε′ costs you at the surface. Reflection R = |(1−√ε*)/(1+√ε*)|² is about 63% for water. In a cavity that power is not lost, but it does reshape the field.

5. Volumetric heating, and why the gradient inverts

This is the practical difference between a microwave and a furnace, and it matters most for solid-state synthesis.

• Furnace: heat enters through the surfaces and flows inward. The surface is always hottest, the core always lags. How fast you can ramp is limited by thermal conductivity and by the stress a gradient puts on the piece.
• Microwave: heat is generated everywhere at once, and the surfaces are the only places it can escape — by convection, and above ~600 °C dominantly by radiation. So the core runs hotter than the surface, the exact reverse of the furnace case.
• Consequences. Good: no thermal lag, so ramp rates of hundreds of K/min, and densification without the usual gradient stress. Bad: the hottest point is the one you cannot see. A surface pyrometer under-reads, sometimes by hundreds of degrees, and interior melting in an apparently intact pellet is a normal failure mode.
• Insulation flattens it. The fibreboard casket around almost every microwave sintering rig is there to cut the surface loss. Less loss at the faces means less gradient, at the cost of hiding the sample even more thoroughly.
• It only holds while d ≲ D_p. If the sample is much thicker than the penetration depth, the source term is itself concentrated near the front face and the profile starts to look conventional again.

6. The other heating mechanisms

Total absorbed power density, electric and magnetic:

Susceptors

Most interesting synthesis targets are transparent cold and absorbing hot — the gap is bridged with a susceptor packed around the charge. Charcoal, graphite and carbon black absorb from room temperature and reach 1000 °C in minutes, though they buffer oxygen. SiC is the standard reusable choice, stable in air to ~1500 °C. Once the charge is hot enough that its own ε″ has risen, it takes over and heats volumetrically; that handover is what "hybrid heating" means.

For La₂O₃ + Fe₂O₃ → LaFeO₃, the two reagents could hardly be more different — compare their rows in the tables below at 25 °C and at 800 °C. The lanthana is transparent throughout; the mixture couples through the iron oxide and through interfacial loss at particle contacts, and the coupling shifts again as the perovskite forms.

Skin depth at 2.45 GHz

7. Summary

• Heating rate follows ε″ and field strength — not ε′, and not tan δ on its own.
• A high ε′ means strong polarisability, so a large possible loss. You only collect it if τ puts the relaxation near your frequency.
• Temperature acts through τ, Δε and σ. Falling ε″ is self-limiting; rising ε″ is runaway.
• D_p says where the heat lands; conductivity and surface loss then decide the gradient.

At 25 °C, 2.45 GHz

At 800 °C, 2.45 GHz

Solids taken as fully dense. A pressed powder compact is mostly air: at 60% packing, ε′ and ε″ both drop by roughly half (Landau–Lifshitz–Looyenga mixing), and D_p doubles.